The Index
Shipped products, working prototypes, tools I run on my own machine, and one research artifact that concluded against me. Nothing here is inflated — if something has no users, it says so.
Most of these exist because of a decision I wrote up first: shipping an offline PWA instead of an App Store build, keeping tools client-side so no data leaves the device, and what a trading journal has to record to be worth keeping.
01 · Freight & AI Automation
Local-first AI that runs inside a real operations team — no cloud, no per-token bill, a human always clicks send.
A luxury super-app study: five product surfaces — retail, stays, cinema, dining, motors — unified under one design system and one faceted search engine. 1,910 items, 63 filters, live facet counts that recompute against every other filter, and shareable URL state. No framework, no build step, no third-party requests; every cover is generated SVG. All catalogue content is fictional demo data.
AI intake, billing and accounts-receivable automation built during my internship at a freight brokerage. Reads carrier messages, drafts the reply, and clears the back office — the AR and billing workflow load, a passing regression suite, entirely on local models.
The HwyHaul systems productized for small brokerages: an AI back-office clerk that reads the billing and AR inbox and drafts every reply on a local model. No API fees, and the data never leaves the building. Founding-client pilots are open.
A flat-fee, student-run campus food delivery concept for SJSU: nearby restaurants on one dark map, delivered by fellow students, with a courier jobs board. Designed and prototyped as an interactive demo, never operated.
02 · Markets & Quant
Built with real engineering discipline and validated adversarially. None of it is trading real money, and I explain exactly why. Background reading: walk-forward testing, how backtests overfit, and R-multiples and expectancy.
A CLI that scans a watchlist for anchored-VWAP, VCP and gap setups, plus a no-lookahead backtest engine — and then the adversarial audit I ran to break my own headline result. It broke. Against risk-matched random entries the signal's own contribution was +0.030R, 95% CI [−0.014, +0.077]. Not a validated edge. Verdict: marginal, don't size up.
An execution engine where the strategy is forty lines and everything else is crash-safety: deterministic order IDs, a SQLite journal claimed before every broker call, linked exits, ownership boundaries, and idempotent reconciliation. It refuses to touch a live broker until 60 closed paper trades show positive realized expectancy — a gate written into the engine, not a policy I have to remember. 160 tests. No live capital.
An R-multiple trade journal that runs in the browser: expectancy, win rate, profit factor and an equity curve, computed on-device. No account, no signup, no backend, no analytics — every trade sits in your own browser storage and never leaves the device. Add it to your Home Screen from Safari and it works with no connection at all.
Trade-journal analytics as a web app: import an R-multiple journal CSV and get expectancy, drawdown, per-setup edge and a plain-English verdict. Analytics on your own history only — no broker credentials, no advice.
A Python toolkit for OHLC breakout and reversal analysis — the earlier research line that fed into the walk-forward harness above. Runs locally, not published as a product.
03 · Startup Prototypes
Working end-to-end on my machine with tests green in CI. No users, no revenue, no public launch yet — that's the honest state.
Raw event footage in, finished 9:16 highlight reel out. Scene detection finds the moments, audio-energy analysis confirms them, ffmpeg renders vertical at social loudness. Zero API keys, zero Python dependencies, plus a local web studio.
A map of reported trash plus a quantified "world fixed" meter — clean a pin, log an eco action, watch kg of waste and CO₂e climb. Privacy designed in at the schema: no user GPS stored, coordinates rounded, hazards are report-only.
A Twitter alternative with the outrage machinery removed: chronological only, no vanity metrics, no quote-posts, no verification hierarchy. Posts are proof of work. Working single-player prototype.
04 · Personal Tools
Small, offline, installable. Built because I needed them daily — free for anyone who wants them. The reasoning behind them is written up too: why I ship an offline PWA instead of an App Store build and why these tools stay client-side.
A daily weight tracker with a 7-day trend line, a goal line and streaks. The tool behind my documented 260→200 lb cut. No account, no server, no tracking — weigh-ins live in your own browser storage, and the app runs fully offline once installed.
A daily execution checklist that resets every morning and tracks a 100% streak. Ship the list, every day. No account, no signup, no server and no tracking — the list is stored in your own browser and works offline. Nothing syncs between devices, by design.
A phone helper for older adults: four huge buttons, guides that give one instruction per screen, a scam checklist that ends in "do not send — call your person," and tap-to-call trusted contacts. High contrast, adjustable text, no accounts.
A zero-dependency security scanner for startup codebases: 24 secret-detection rules, dependency CVEs via OSV.dev, 38 misconfiguration rules, license risk, git-history leaks, SBOM export and an 18-point readiness score. MIT licensed.
My personal operations dashboard: a local Next.js app over SQLite that keeps internship applications, deadlines, trackers and daily execution in one place. Runs on localhost, for an audience of one.
Built and maintain sventclub.org, the front door for the Silicon Valley Entrepreneurship Club — events, sponsors and the member funnel. Part of the push to make the club's events and membership easier to run.
Install the web apps on your phone
Android or desktop: open the browser menu or the install icon in the address bar and choose Install app. Everything is local — no account, no signup, no server, no analytics, no cookies. The honest tradeoff: nothing syncs between devices, nothing is backed up, and clearing your browser data or deleting the app erases your entries — so use each app's export button if the history matters. Full install steps: TradeLog · Cut · Fire.
Not looking to install anything? The same math runs in the browser with nothing to add to your home screen: six trader risk calculators (position sizing, R-multiples, expectancy, Monte Carlo risk of ruin, Kelly and break-even win rate) and six freight-brokerage calculators (rate per mile, deadhead, margin, detention, fuel surcharge and break-even load).
05 · Brand & Commerce
The commercial side: a streetwear label, digital products, and services — all designed, built and shipped solo.
A bilingual Arabic–English streetwear label built in public. Drop 001 is six pieces with a full brand kit and print-ready artwork; the storefront is live and taking waitlist signups ahead of fulfilment going online.
Automation software, playbooks and premium template packs from $19 — including the Swing Trading Operating System course and the FreightDesk install. Sixteen products plus bundles, all written and designed by me.
A single-product apparel concept: 450+ GSM heavyweight hoodies, minimal design, no loud logos. A full brand and landing-page build exploring how far one product can carry a label.
Twelve black-and-gold template packs: investor pitch deck, a 60-layout slide library, client proposal and SOW, a freight-broker ops pack, and brand guidelines. Designed to be used, not admired.
A productized service: a complete LinkedIn profile rewrite in 48 hours — recruiter-stopping headline, rebuilt About section, metric-first bullets and keyword mapping.
My brand-collaboration one-pager: audience stats, the 260→200 lb transformation story, and the partnership options I offer. The page I send instead of a PDF.
06 · Research
CS education research at the SJSU CSEd Research Lab under Dr. Ethel Tshukudu — mixed-methods studies on how students learn to program.
Co-authored a mixed-methods study with 60 participants on how bilingual programming environments lower the barrier for non-native English speakers learning CS. SIGCSE TS 2026, ACM Digital Library.
A free 8-week bilingual intro-CS program for English-learner high schoolers, built directly out of the bilingual-coding research above. Curriculum, volunteer handbook and partnership kit are written and public; a signed school partner and a first cohort are next — stated honestly on the page itself.
Built through IBM SkillsBuild: an IBM Watson–powered sentiment engine scoring equities from market news, built end to end. A learning project; it was never run in production, so there are no audited performance figures to quote.
A note on repos. Several of these live in private repositories — the trading work, the startup prototypes, and my personal ops dashboard. I'm happy to walk through any of them on a call or share access on request. Client and employer work is not published at all.
FAQ
Four installable offline web apps: TradeLog, an R-multiple trading journal; Cut, a weight and body-recomposition tracker; Fire, a daily execution-ritual checklist; and GRAMPA, a big-button phone helper for older adults with step-by-step guides and scam protection. The site also hosts six free trader risk calculators, six freight-brokerage calculators, a visa planner and an internship application tracker. All of them are free with no signup.
Neither. There is no signup, no login and no backend. Once installed, the apps work fully offline and store data in your device's local browser storage. The honest tradeoff: nothing syncs between devices and nothing is backed up, so clearing your browser data or deleting the app erases your entries.
Open the app in Safari, tap the Share button, scroll down and tap Add to Home Screen, then tap Add. It has to be Safari — iOS does not let Chrome, Firefox or an in-app browser install a web app. Once added, it launches full-screen with no browser chrome and runs with no connection at all.
On Android, open the Chrome menu and choose Install app or Add to Home screen, then confirm. On a desktop, click the install icon in the Chrome or Edge address bar, or open the three-dot menu and choose Install, then confirm. Chrome may also offer an install banner on its own.
Everything stays in your own browser's local storage on your own device. There is no telemetry, no analytics on the app data, and no server to send it to, because none of these apps have a backend. The full policy is on the privacy page.
It is deleted permanently and cannot be recovered, because there is no copy anywhere else. This is the direct cost of the no-account design. If the history matters to you, use the app's export button regularly and keep the file somewhere you back up.
They are free with no paid tier, no trial, no ads and no upsell inside the app itself. The paid products on this site are separate: a trading course, template packs and automation tooling in the store. Installing an app never asks for payment or a card.
Yes. Each app is a single small HTML file with no framework, no build step and no third-party requests, so the first load is quick and every load after that is served from the device cache. Once installed, connection speed stops mattering entirely because nothing is fetched.
No. TradeLog is a record-keeping and arithmetic tool — it computes R-multiples, win rate, expectancy and an equity curve from trades you type in yourself. It gives no signals, no recommendations and no forecasts. Trading carries substantial risk of loss, including your entire capital.
The web apps are plain HTML, CSS and JavaScript with no build step, so the full source is readable with View Source on any of them. Several of the larger projects live in private repositories — the trading systems, the startup prototypes and the personal ops dashboard — and I am happy to walk through those on a call.
TradeLog, the swing screener and every trading-related tool linked here are educational and record-keeping tools only. Nothing on this page is financial, investment, tax or legal advice, or a recommendation to buy or sell any security. Trading carries substantial risk of loss, including your entire capital. Backtested and walk-forward figures shown here are hypothetical, do not represent live trading, and past performance does not guarantee future results.